[Feature] Add sandbox

[Yunnglin]

Start server first, you can run remote or local:

from evalscope.sandbox import DockerSandboxServer

server = DockerSandboxServer()
server.run()

• Usage1:

from evalscope.sandbox import DockerSandboxClient

# Using the client as a context manager
with DockerSandboxClient("http://localhost:8000") as client:
    # Execute code that imports libraries
    setup_result = client.execute_code(dedent("""
    import pandas as pd
    import numpy as np
    
    # Create a sample dataframe
    data = {
        'Name': ['Alice', 'Bob', 'Charlie', 'David'],
        'Age': [25, 30, 35, 40],
        'Score': [85.5, 90.0, 78.5, 92.5]
    }
    
    df = pd.DataFrame(data)
    df.to_csv('/sandbox/data.csv', index=False)
    print("Created sample data")
    """))
    
    print(setup_result)
    
    # Read the generated CSV
    file_content = client.read_file("/sandbox/data.csv")
    print("CSV file contents:")
    print(file_content)
    
    # Analyze the data
    analysis_result = client.execute_code(dedent("""
    import pandas as pd
    import numpy as np
    
    # Read the data
    df = pd.read_csv('/sandbox/data.csv')
    
    # Perform some analysis
    print(f"Average age: {df['Age'].mean()}")
    print(f"Average score: {df['Score'].mean()}")
    print(f"Correlation between Age and Score: {df['Age'].corr(df['Score'])}")
    """))
    
    print("Analysis results:")
    print(analysis_result)

• Usage 2:

from evalscope.sandbox import DockerSandboxClient, DockerContainerConfig

# Initialize client
client = DockerSandboxClient("http://localhost:8000")

# Create a container
container_id = client.create_container(
    DockerContainerConfig(
        memory_limit="2g",
    )
)

try:
    # Execute Python code
    result = client.execute_code(dedent("""
    import numpy as np
    import matplotlib.pyplot as plt
    import io
    import base64
    
    # Generate some data
    x = np.linspace(0, 10, 100)
    y = np.sin(x)
    
    # Create a plot
    plt.figure(figsize=(8, 4))
    plt.plot(x, y)
    plt.title('Sine Wave')
    plt.grid(True)
    
    # Save plot to a buffer
    buf = io.BytesIO()
    plt.savefig(buf, format='png')
    buf.seek(0)
    
    # Print the base64 encoded image
    print(base64.b64encode(buf.read()).decode('utf-8'))
    """))

    print(f"Execution output: {result['output']}")  # Show beginning of output

    # Execute a shell command
    cmd_result = client.execute_command(["ls", "-la", "/sandbox"])
    print(f"Command output: {cmd_result['output']}")
    
    # Write a file
    client.write_file("/sandbox/test.txt", "Hello from the sandbox!")
    
    # Read a file
    content = client.read_file("/sandbox/test.txt")
    print(f"File content: {content}")
    
finally:
    # Delete the container
    client.delete_container()